In this paper, Back Propagation (BP) neural network and Support Vector Machine (SVM) were used to identify the high-strength bolts looseness damage based on piezoelectric wave method. The effects of algorithms and damage indexes and sample sizes for the identification of high-strength bolts looseness damage were studied. The results showed that the BP neural network recognition accuracies of wavelet packet node energy change rate (WPNECR) damage index in “small sample” and “large sample” were 75% and 100%, respectively. And the BP neural network recognition accuracies of wavelet packet node energy entropy (WPNEE) damage index in “small sample” and “large sample” were 92.7% and 100% respectively. In addition, the SVM recognition accuracies of wavelet packet node energy entropy damage index in “small sample” and “large sample” were both 100%. It shows that wavelet packet node energy entropy as damage index can better reflect the high-strength bolts looseness damage characteristics than wavelet packet node energy change rate. So, BP neural network and SVM can better identify the high-strength bolts looseness damage, but SVM has more advantages in the case of “small samples”.